Assessing nonresponse bias at follow-up in a large prospective cohort of relatively young and mobile military service members

Seattle Epidemiologic Research and Information Center, Department of Veterans Affairs Puget Sound Health Care System, Seattle, WA, USA.
BMC Medical Research Methodology (Impact Factor: 2.27). 10/2010; 10(1):99. DOI: 10.1186/1471-2288-10-99
Source: PubMed


Nonresponse bias in a longitudinal study could affect the magnitude and direction of measures of association. We identified sociodemographic, behavioral, military, and health-related predictors of response to the first follow-up questionnaire in a large military cohort and assessed the extent to which nonresponse biased measures of association.
Data are from the baseline and first follow-up survey of the Millennium Cohort Study. Seventy-six thousand, seven hundred and seventy-five eligible individuals completed the baseline survey and were presumed alive at the time of follow-up; of these, 54,960 (71.6%) completed the first follow-up survey. Logistic regression models were used to calculate inverse probability weights using propensity scores.
Characteristics associated with a greater probability of response included female gender, older age, higher education level, officer rank, active-duty status, and a self-reported history of military exposures. Ever smokers, those with a history of chronic alcohol consumption or a major depressive disorder, and those separated from the military at follow-up had a lower probability of response. Nonresponse to the follow-up questionnaire did not result in appreciable bias; bias was greatest in subgroups with small numbers.
These findings suggest that prospective analyses from this cohort are not substantially biased by non-response at the first follow-up assessment.

Download full-text


Available from: Gary D Gackstetter, Oct 04, 2015
18 Reads
  • Source
    • "Even though only few in the target population participated, the main effect of non-participation was a loss of precision in stratum-specific estimates [26]. Similar conclusions were drawn in a prospective cohort study, in which it was found that a prospective analysis in a cohort of relatively young, highly mobile, adult military personnel was not substantially biased by non-response at the first follow-up after four years [27]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: A randomized intervention study, "Preventive consultations for 20- to 40-year-old young adults", investigated whether preventive consultations with a general practitioner could help young adults with multiple psychosocial and lifestyle problems to change health behavior. To optimize the response rate of questionnaires at 1 year post-intervention, the non-responders were reminded by telephone. The aim of this study was to examine potential selection bias induced by non-response by comparing responder and non-responder populations at baseline, and to examine the impact on outcomes by comparing initial respondents to respondents after telephone reminding.
    BMC Research Notes 09/2014; 7(1):632. DOI:10.1186/1756-0500-7-632
  • Source
    • "The study population consisted of the entire study samples for Panels 2 and 3 of the Millennium Cohort Study, except those who refused to participate. Panel 1 was enrolled during 2001-2003 and consisted of 77,047 (30%) participants from a probability-based sample of 256,400 military personnel who had at least 1 year of service as of October 1, 2000; this panel has previously been studied in terms of nonresponse to enrollment [24] and follow-up [9] in relation to health status. For Panel 2, 150,000 randomly selected military personnel with 1-2 years of service as of October 2003 were invited. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Longitudinal cohort studies are highly valued in epidemiologic research for their ability to establish exposure-disease associations through known temporal sequences. A major challenge in cohort studies is recruiting individuals representative of the targeted sample population to ensure the generalizability of the study's findings. We evaluated nearly 350,000 invited subjects (from 2004-2008) of the Millennium Cohort Study, a prospective cohort study of the health of US military personnel, for factors prior to invitation associated with study enrollment. Multivariable logistic regression was utilized, adjusting for demographic and other confounders, to determine the associations between both deployment experience and prior healthcare utilization with enrollment into the study. Study enrollment was significantly greater among those who deployed prior to and/or during the enrollment cycles or had at least one outpatient visit in the 12 months prior to invitation. Mental disorders and hospitalization for more than two days within the past year were associated with reduced odds of enrollment. These findings suggest differential enrollment by deployment experience and health status, and may help guide recruitment efforts in future studies.
    BMC Medical Research Methodology 07/2013; 13(1):90. DOI:10.1186/1471-2288-13-90 · 2.27 Impact Factor
  • Source
    • "More specifically, several studies have been undertaken on the non-response to health interview surveys (HIS). Previous studies find that respondents have a higher socio-economic status and that they report a better subjective health, lower healthcare use, and healthier lifestyle behaviour than non-respondents [19-28]. However, other studies find higher healthcare use for respondents or better health status for non-respondents [29-33]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: Unit non-response occurs in sample surveys when a target subject does not respond to a survey. Potential implications are decreased power, increased standard error, and non-response bias. The objective of this study was to assess the factors associated with participation in a written survey (MSHS) of subjects who had previously participated in the Swiss Health Survey (SHS) and to evaluate to what extent non-participation could impact the estimation of various MSHS health outcomes. Methods: Multivariate logistic regression was used to assess the factors associated with MSHS participation (n=14,393) by eligible SHS participants (n=17,931). Crude participation rates and the adjusted odds ratios of participation (OR) were reported. In order to report potential bias in MSHS outcomes, the average age-standardized and sex-specific outcome values in non-participants were predicted based on several different linear regression models which had been previously fitted on MSHS participants. Results: Adjusting for all other variables, women (OR=1.63) as compared with men, subjects with a secondary (OR=1.48) or tertiary education (OR=1.76) as compared with those with primary education, white-collar workers (high level non-manual workers OR=1.29, medium and low level non-manual workers OR=1.26 and OR=1.25 respectively) as compared with unskilled manual workers, Swiss nationals (OR=1.60) as compared to non-Swiss, and subjects with very good or good self-rated health (OR=1.35) were more likely to participate in the MSHS. People who work full-time were less likely to participate than those without paid work (OR=0.76). There were no statistically significant differences in the likelihood of participation between rural and urban areas, different geographic regions of Switzerland and household income quartiles. Except for myocardial infarction, all age-standardized and sex-specific average outcomes (influenza vaccination, arthrosis, osteoporosis, high blood pressure, depression, mastery, and sense of coherence) were significantly different between MSHS non-participants and participants. Conclusions: Subjects who participated in the MSHS had a higher socio-economic status, reported a better subjective health, and were more likely to be Swiss nationals. Small to moderate bias was found for most age-adjusted and sex-specific average outcomes. Consequently, these MSHS outcomes should be used and interpreted with care.
    BMC Public Health 02/2013; 13(1). DOI:10.1186/1471-2458-13-167 · 2.26 Impact Factor
Show more